LAUSR

Support vector machines (SVMs) based on fuzzy theory have attracted widespread attentions in pattern recognition and machine learning. However, these SVMs have some limitation in dealing with some classification problems with fuzzy outputs, which results in the ignorance of the fuzziness of fuzzy outputs. Motivated by this, the possibility and necessity of fuzziness of fuzzy outputs are discussed, and the dynamic partitioning methods of these fuzzy output training samples are demonstrated based on credibility measure. Then, the corresponding dynamic credibility support vector machines based on fuzzy outputs are established, and the feasibility and effectiveness of credibility SVMs are shown by experimental results.